Significant discrepancies had been identified by many research groups world wide between calculated and measured neutron doses from the ⁷Li(p,n) accelerator based neutron source, and therefore precise characterization of the source was needed. In this work neutron spectra from the ⁷Li(p,n) source were measured with a ³He ion chamber in the incident proton energy range from 1.95 to 2.3 MeV. The ³He detector is hypersensitive to slow neutrons, so a time-of-flight based slow neutron rejection acquisition system was built and tested. The system is based on an electrostatic proton chopper and an acquisition system working on coincidence mode. The response function of the ³He was extended down to 30 keV neutron energy and the collected neutron spectra were unfolded using two methods: van Cittert iterative algorithm with Jansson constraint, and a regularized constrained inversion. Theoretical neutron spectra emitted by the ⁷Li(p,n) source were calculated and compared with experimental unfolded spectra, as well as with results of the Monte Carlo simulations of the lithium target assembly and the walls of the experimental area. Using fluence to kerma conversion coefficients, the neutron dose was calculated and compared with results obtained from an independent experiment using the microdosimetric technique with a tissue equivalent proportional counter. Total neutron yield of the ⁷Li(p,n) reaction was measured using induced activity of ⁷Be. Results showed a negative energy offset of the incident proton beam between 50 and 58 keV with respect to the generating voltmeter indication of the accelerator terminal. Shapes of the measured neutron spectra showed significant moderation originating from neutron scattering on the lithium target assembly and walls of the experimental area. When accounting for this offset, neutron yields showed an agreement with calculated values within 22% for 1.95 MeV and within approximately 7% for higher proton energies. / Thesis / Doctor of Philosophy (PhD)